Running the FETs at lower Vds will run them over a more linear portion of their characteristics but you'll run into higher noise from the lower signal level. So there's a trade-off - of course lower noise opamps than 5534s can help offset this. I guess it depends whether lowest noise or lowest distortion is your goal - if the output is to be fed to an FFT, probably the higher noise isn't too much of an issue as you can always average away noise, given enough acquisitions (time) 
I have looked at the 4402 schematic.
It is not the based on a state variable filter.
4402B Service Manuals
There are a few interesting things that can be found in this schematic- It uses the LM318 with a discrete bipolar output stage
- The oscillator level is 10V, 20Vpp
- It uses the AD633 as part of the gain control
- The actual gain control element is the bias current of bipolar output stage
- It uses a signal from the resonance element as feedback to the gain control element
The schematic was not that easy to figure out, I had to redraw the circuit.
It's drawn in kicad. It had a few export alternatives, I choose the SVG format. I have now exported it in PS and converted it to pdf, as the request.
View attachment KH4402-clone.pdf
The schematic is not the complete 4402,
- The power supply is missing
- The output stage is missing
- All switching circuitry is replaced with a general R or C
I will probablay have missed some more things, comments are welcome.
I have been thinking some more to this. I think it is very difficult to improve this design 20dB with some more type of compensation. I think CR132 was the final improvement, it just looks like that.
Lets look at the conceptual differences
- 4402 uses 20Vpp swing
- 4402 has another type of level detector
- 4402 switches 2*R and 1*C for frequency control
- 4402 uses the JBT bias for gain control
Anyway, my belief is that the quality of an oscillator is proportional to the Q of the resonance element. My subjective view is that the SVF has highest Q, and therefore the architecture for the best performance
View attachment KH4402-clone-svg.zip
Another try sorry for your inconvenience. The pdf is packed into a zip archive.
Another try sorry for your inconvenience. The pdf is packed into a zip archive.
Any oscillator with stable amplitude has, by definition, a Q of exactly infinite; the AGC leveling loop is there to adjust this. With this respect, there are no drastic difference between the various oscillator/filter topologies.
The advantage of the SVF relates to the ease of adjusting frequency, and the inherent low-pass action which reduces distortion from level detector ripple and multiplier.
There is a 4402 manual around which has the schematics added; can't find the link right now...
Samuel
I have the 4402A and 4402B SM and schematics.
I can't post them but I can email them out if anyone is interested.
David.
Last edited:
Well, it is my RF background that gives me these beleives. For old mobile systems like NMT, you needed an oscillator that had very low noise at 25kHz from the carrier and a very low noise floor. An easy way to acheive this, was to use ceramic resonators with high Q.
I need some more explanation to understand your argument.
The oscillator noise is also present in the circuit, with a higher Q in the resonance circuit, my beleif is that the noise would be less, since the noise is filtered by the resonator. This is like 1/f noise. But it might be of no interest in audio.
Then we have the noise floor and harmonic distorsion, with good Q was it easier to get a good noise floor in the RF VCOs.
Anyway, I still beleive the SVF is the better arcitecture. If you are building a THD analyzer, is the SVF used twice, one as source and once as the notch filter. I am a strong beleiver in the reuse of circuit topologies.
Minor schematic correction
Hi,
When I was looking at the schematic, was one of the most important wires missing. I am sorry that I hadn't checked that earlier.
It was the tap from the RC-resonator generating the oscillation.
View attachment 4402B-clone-rev01-svg.zip
Hi,
When I was looking at the schematic, was one of the most important wires missing. I am sorry that I hadn't checked that earlier.
It was the tap from the RC-resonator generating the oscillation.
View attachment 4402B-clone-rev01-svg.zip
Hello again,
One simple question,
I have looked at the schematics of the 4402, 4400 and 4500. The first two use LM318 in the oscillating circuit. The 4500 uses the NE5534 as oscillating circuit.
What is the difference in figure of merit between LM318 and NE5534?
Are there new OP AMPS, with better performance, that actually have been used in this type of circuitry?
One simple question,
I have looked at the schematics of the 4402, 4400 and 4500. The first two use LM318 in the oscillating circuit. The 4500 uses the NE5534 as oscillating circuit.
What is the difference in figure of merit between LM318 and NE5534?
Are there new OP AMPS, with better performance, that actually have been used in this type of circuitry?
The LM4562 is a good modern op amp for these applications. It does, however, have higher input current noise than the 5534, so, depending on the impedance levels in the circuit, it could be a little noiser (in high impedance circuits).
After 30 years, it is remarkable how very good the 5534 still stands up in regard to measurable THD for instrumentation purposes.
Cheers,
Bob
The 5534 is used in the output stage of the Boonton 1120 THD Analyzer. In later versions it is used as a driver for higher output power in the 1120.
Further to Bob's idea of the distortion magnifier :
http://www.diyaudio.com/forums/soli...ordell-distortion-analyser-2.html#post2003882
Here is an interesting discussion at EDN on something similar.
Measure an amplifier's THD without external filters - 2012-01-05 08:00:00 | EDN
Patrick
http://www.diyaudio.com/forums/soli...ordell-distortion-analyser-2.html#post2003882
Here is an interesting discussion at EDN on something similar.
Measure an amplifier's THD without external filters - 2012-01-05 08:00:00 | EDN
Patrick
I strongly suggest you look at the LT1468. It has nearly the same distortion performance as the LM4562, but *much* lower current noise. This makes it more suitable for the impedance levels found in typical oscillator implementations. The somewhat higher voltage noise doesn't matter much if the oscillator level is carefully chosen.
If you go with the NE5534, there are some compensation tricks which I believe will lower distortion (didn't verify this yet). See Tektronix SG505 or AP System 1 service manual.
That measurement method is old news--used extensively by Walt Jung and myself (see www.diyaudio.com/forums/solid-state/129202-amazing-opamp-measurement-shoot-out.html).
Samuel
Hi Samuel,
I thought you might be able to shed some light on an effect with a SVF/SVO.
I working on a SVF/SVO design using the LT1468. I don't think this effect is particular to this op amp but to an un-damped, SVF. FB resistor removed from the BP section.
For two adjacent ranges with overlapped Fr, two different RC pair have a different gain at resonance.
An example is 2k ohm and 10nF, 20k ohm and 1nF. Each pair have the same Fr but the 2k/10n pair has a Fr gain of 58dB and the 20k/1n pair has a gain of 46dB.
Input is through a resistor into the inverting node of the inverter op amp of the Hp section.
The resistor is set equal to the FB resistor of the inverter amp HP section.
The input signal is tuned to the FR of the filter. The input level is adjusted to produce a same output level for each frequency tested and the gain evaluated from Vout/Vin. The output is taken from the BP section.
One thing that stands out is that the current into the virtual nodes of the integrators is 10 times different from one RC set to the other and the virtual ground at the said node is maintained by the op amp's OLG. It stands to reason that this effect of gain difference may be related to amount of available op amp OLG.
David.
I thought you might be able to shed some light on an effect with a SVF/SVO.
I working on a SVF/SVO design using the LT1468. I don't think this effect is particular to this op amp but to an un-damped, SVF. FB resistor removed from the BP section.
For two adjacent ranges with overlapped Fr, two different RC pair have a different gain at resonance.
An example is 2k ohm and 10nF, 20k ohm and 1nF. Each pair have the same Fr but the 2k/10n pair has a Fr gain of 58dB and the 20k/1n pair has a gain of 46dB.
Input is through a resistor into the inverting node of the inverter op amp of the Hp section.
The resistor is set equal to the FB resistor of the inverter amp HP section.
The input signal is tuned to the FR of the filter. The input level is adjusted to produce a same output level for each frequency tested and the gain evaluated from Vout/Vin. The output is taken from the BP section.
One thing that stands out is that the current into the virtual nodes of the integrators is 10 times different from one RC set to the other and the virtual ground at the said node is maintained by the op amp's OLG. It stands to reason that this effect of gain difference may be related to amount of available op amp OLG.
David.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.